Jeff Pressing COGNITIVE ISOMORPHISMS BETWEEN PITCH AND RHYTHM IN WORLD MUSICS: WEST AFRICA, THE BALKANS AND WESTERN TONALITY This Paper compares sonve diverse musical phenomena in the light of their iAdrWINE structural similarities. Specifically, a number of common covlic Sandable ;2 Bch and thythm are tound to be isomorphic, and to he under, LasMable in terms of the principles of mathematical graun theory, Beceee ‘uch pitch and ciythmy patterns are the products of hutnan musica thinking, 1 Sa ie reationsnips between them cogntive isomorphisms, By ths phrase 1 Heim amcan 10 suREest thar detailed cognitive mods of sucn patterns se being presented—eather, tha the observed strtial similwitin ee suffi ciently compelling, and their relation to musical perception and training suffie Gently direct, to justly the hypothesis that they may result fran general Seiad hee ee Lhe Basie idea of group-theoretic interpretationtie nee orginal here (eg., ce Cassiter, Balzano), and in view of rhe applicability of Froup theory to such diverse areas as cognitive and Gestalt Psychology, quan- aT Eitsics erosallography, change-tinging, general sytere theory, Scottish eying decorative ceramis design, kot theory and ile anise seinen Stubhikow and Koprsk, Ralzano), cursory memion ein made ‘of elated formulations outside the field of music, 1 akin with the perceptual space of interest wil be defined in an abstract ampceneral orm. Consider a one-dimensional array of lace tc, potentially Unbounded in extent, with an ‘equivalence operates” ‘operating on the array so that sites L units distant from one another are it co example, if these lattice sites are , then Lis {2 and the equivalence operator ig the octave.) If M objects (M< L} pater gitibuted armong any b adjacent lative sites, we vbonee “unit cell" Paucar wich may be replicated over the entire late, {hoe example, the white Bods ofthe piano form a unit cell pattern) These en cn may operate as a Kind of underlying reference pattern for athet ‘ccupancy paiterns overlaid on Preface, Because ofthe equivalence operaion the een no longer most aniy represented as strictly one-dimensional. fn tech Shephard has shown, ting pitch as his domain of soneern, that @ double: structure based onne COGNITIVE ISOMORPHISMS two independent variables of pitch height and (one chroma provides a very food geometrical representation of many properties of sucha constrained lattice. Now, to the extent that this kind of scheme corresponds to a general cog- nitive phenomenon, we should expect to encounter equivalent unit cell struc- tures, for given M and L, in different but equivalently constrained perceptual domains (e.g., musical chythm, pitch, one-dimensional visual patterns, poetic meter). We might also expect to find predilections for specific values of M or L, or for some particular relation between them, and to be able 10 formulate general principles of sequential organization of naturally occurring unit cells. To investigate this in more detail, consider the phenomenon of music, with L = 12, Here it is possible to compare the rhythmic organization of much West African drum ensemble music (that in 12/8 meter) with the pitch organization in Western art music (12 semitones per ociave), In the above terminology, the respective lattices consist of discrete points along the axes of time and frequency (notes), the respective equivalence operators are ihe regularly recurring rhythmic cycle and the octave, and the respective unit cells, are the repeating hell patterns (see below) and the Western scale choices. The unit cells in the to musics may be considered to function analogously or be equivalently constrained, in thatthe bell pattern serves asa time-line reference arid for overlaid rhythmic patterns, and the Western scale serves as a pitch reference grid for the melodie choices of tonality or modality. Some background information on West African drum ensembles is prob- ably appropriate at this point. Much of this music consists of overlays of various rhythmic strata. For example, in snusic of the Ewe people of Ghana, a stratum of high-pitched instruments, typically the drum kagan, the rattle axaise, and the double bell gankogni, along with clapping, play virtually uunvarying patterns which form a stable basis over which the other drums kidi, sogo and avsimevu play more variable patterns. Kid? and sogo may be con- sidered to form a second stratum, and norreally follow changes in ausimevu, the master drum, which is itself a third stratum, Singing and dancing go on simultaneously. The prime responsibility for time-keeping is acknowledged to lie with the bell player (Jones 1959; Locke; C. K. Ladzekpo), and all patterns are fundamentally conceived in relation to the bell, although perhaps also in relation to a basic pulse and other interlocking drum parts in the ensemble. The rattle axaise often plays the same pattern as the bell, or a variant based on it. Kagan and the clapping usually serve to reinforce the division of the bell pattern into numbers of equal parts. Because of the special acknowledged role Of the bell, and its timbral differentiation from the other instruments (itis the only metallophone), it seems valid to consider it as an underlying time-tine ith only occasional references to other parts. Where bell patterns do not so function, as in the gamamle of Ewe music (many bells), this isolation principle may fail to hold, The word ‘may’ is appropriate here, as even when there is no bell, another instrument (or sometimes clapping) typically performs its time- Tine organizing function with a similar or identical rhythmic pattern (for ‘example, in Yoruba sacred music of Ekiti, this isthe talking drum kdnango— of. King), "Among the Ewe, whose music is probably the best understood of this 39STUDIES IN MUSIC region, one bell pattern predominates, namely J J DJ 1 J | . 1s 128 suructure may be represented in terms of note durations (strictly speaking, time between bell strokes, as in some other patterns the bell is damped) as 2242221. Furthermore, this same cyclic structure is preserved with all possible starting. Points with respect to the pitch domain in modal music of the medieval Euro. Pean period, natural minor tonality of the classical era, impressionism and centric 20th century music from jazz to the symphonies of Roy Harris, and with respect to rhythm in the time-line patterns of other tradional musics of West African origin.* So al the properties of the major scale, such as the cycle of fifths, or isomorphic tetrachord construction, may potentially be found mirrored in these African time-lines, A comparison of modal treatments of pitch and rhythm for this M = WL = 12 case is given in detail in Table 1.* By ‘mode’ a unit cell structure with: specified starting point is meant. Examples of Western music with the in cated pitch structures have not been ‘given as it is assumed the readcr is familiar with this repertoire. Included are olf the time-lines 1 have been able to glean from a search of the literature and from Personal study with several West African drummers which feature 7 strokes ina cycle of 12 beats. Modes of the major scale structure overwhelmingly predominate, and examples of all ‘rhythmic modes’ except phrygian have been found. Pattern & is different, being a mode of the ascending melodic minor scale, but like the other patterns, is made up of only the elements | and 2. Pattern 9 is from a juvenile song and does not correspond to acommon Western scale. ‘Why is the 2212221 structure so special in both pitch and rhythm domains? In pitch representation, this is an issue with a long and hoary tradition of explanations and diatribes; before delving into this, it will be useful to look at other representative 12/8 bell paiterns in the music of this region, In Table Ia number of these are displayed. (Two of the Palterns—numbers § and 6—are from southern Africa, but they seem closely related.) Among these, pattern 3 (22323) is considered Particularly important (Jones 1989, 1965; Ekwueme) due to its high incidence and its ability to readily generate, by simple rules of trans- formation, almost all the other 12/8 Patterns in the table. For example, the 4th (23223), Sth (32322) aud 6th (32232) patterns follow from it bya shift of modal Starting point, and pattern 4 is sometimes considered equally fundamental with pattern 3, due to high incidence. Other Patterns may be seen to be related. to pattern 3 by the transformation rules of element fusion or fission that corre- spond to the common musieai processes of elsion or fillig-in, Thus pattern 1 may be derived from 22323 by fusion of ihe two 2's and cyclic permutation, Pattern 9 by fission of the second 3-element into 1+ 2, pattern 7 by fusion of the Wo 2's and fission of one 3, and the basic pattern 1 of Table 1 (2212221) by splitting both 3-elements, respectively into 1+2and2+1. Pattern 10 is alsoso understandable, as a composite of (he Standard 7-stroke pattern and a steady 4-pulse, Strokes |, 3, 6 and 8 of this Pattern are made on the lower bell of the gankogui, and no others are, which Supports this interpretation.? Patterns & ant 2, which are not particularly common, may only be derived from 27394 by allowing clement permutation as a transformation operation. This isa mons * Tables I-V are tobe found athe end ofthis ance 0COGNITIVE ISOMORPHISMS drastic type of transformation, perhaps particularly plausible where rhythmic organization is additive. The pattern 22323 is, in the pitch domain, none other than the anheritonic pentatonic scale CDEGA which has such far-reaching distribution in world music. In West Africa, forms of this pentatonic also occur widely (Nketia 1963a), along with a number of others, notably CEFGB and CEFEGB (C. K. Ladzekpo). Intonation naturally differs from Western norms. West African pitch organization when not pentatonic is most often heptatonic, with non- tempered versions of modes of the major scale used (Nketia 1963a).’ In view of the transformation rules, the isomorphisms generated from this one basic pattern seem far-reaching, and correspondences for the others are alluring: II=1 is the minor seventh chord, H—2 is the dominant seventh chord, and I~ 10s a kind of blues scale. Only the two Ashanti patterns I~ 7 and 11-9 do not have common pitch correlates." Further examples of forms of 22323 in the shythm domain may be found in ‘other musics with consistent unequal subdivisions of the 12-cycle. Macedonian ‘and Bulgarian folk songs in 12 are known to use the following rhythmic sub- divisions: 32232, 32322, 22323, 23223, 32223 (Singer, Kremenliew). Here 32232 is pechaps most common, occurring primarily in western Macedonia. Once again the transformations of applying different starting points and (in the last ‘case) element permutation can simply account for different forms of the same basic pattern, Interestingly, of the five cyclically permuted forms of 22323 possible, all occur here except 23232, which is the only mode of the five to be Absent from the African music of Table II as well. Itis also a fairly rare mode Of this formation in pitch space. This may be due to the fact that only this pattern neither starts nor ends with a 3. This will be discussed further in a more ‘general context below. ‘Another one-dimensional space which bears some relation to the model used here is that defined by poetic meter. For example, the structure of traditional Japanese verse (Haiku, Waka, Reng) is dominated by constructions of alter- nating $°s and 7's, which allows partitioning which bears resemblance to patterns already found here. And in fact the work of Halle and Keyser on the analysis of English stress patterns shares a number of general features with the present model. Their analysis consisted of the identification of abstract metrical patterns (usually made up of 1 or 2 kinds of elements), followed by the discovery of rules of correspondence which allow a particular example to be seen as an instance of an abstract pattern. However, 1 do not wish t0 go more deeply into this now, partly because poetic meter is often not metric in the same recurring way that music is (with words, the perceptual lattice can be stretched like a rubber membrane) and hence the details of rules of transfor mation differ. ‘A transformation rule not mentioned so far, but of great generality, isthe idea of complementation. That is, if a pattern is perceptually fundamental, so ‘must be its complement, which psychologically merely corresponds to a figure ‘ground reversal. Complement here is used in the sense of set theory, that is, the Complement of a set of all (unit cel) elements not in S. The imporxance of the Complement isa time-honoured principle of Gestalt psychology (Koffka), and although it's primarily associated with vision in that literature, it has recently 4STUDIES IN MUSIC been used as # basis for understanding the perception of rhythmic patterns as well (Preusser 1 af., 1970; Garner 1974; Oshinsky and Handel 1978). It és quite clear that complementation and figure-ground relations are Sritical in the performance and perception of West African drum ensemble music, A simple example is the very common Ewe kagan patterns 5-8 of Table V, which is conceived in complementary relation to a basic 4 pulsation, ‘And often, as for example on the Ewe drums kidi and sogo, an accented pat. tern of bounce strokes is filled in completely with muted strokes made by pressing the sticks to the drum head, to help maintain time precision: the mute Strokes are then the complement of the bounce stroke pattern, and though not heard prominently in performance, they constitute a strong kinesthetic Dresence, Likewise, in other hand percussion rechniques where the pattern accents are played completely with one hand, the other will fll in the pulses to create 2 complementary motor image (e-g., some versions of Latin conga tech. nique). Inthe context of pitch, the pentatonic and major scale structures are com plements to each other, 28 the white and black key groupings of the piano key- board are known to demonstrate, Hence |) . J Jf} is not only derivable from |J J J. | J. | by fission ofthe two S-clements, but per- haps more fundamentally ‘by’ complementation or figure-ground reversal as well. This sort of derivation emphasizes the holistic nature of the patterns, rather than the values of or interrelations between their components. Picking unambiguously between these two strategies of explanation is hampered by theit equal success, though complementation is perhaps more conceptually economical ‘An independent musical tradition sheds some light on this issue. In 20th cen- tury music by Schonberg, Webern, Babbitt and others, sets of pitches and their complements play a prominent role. In the theory supporting this music ef. €.8~ Forte) is shown that any set of notes and its complement have impor. {ant structural similarities which facilitate their integrated usage,’ This con ‘monality of lattice structure-based complements has important ramifications ulside of music as well, as for example in the highly important sing model (sing, T. Hil) of statistical physics, where the symmetry of the two-particle distribution function in ferromagnetic or lattice gas systems at complementary densities, based on particlesiole symmetry, is proven for any number of dimensions and may be shown to be equivalent to and hisioricaly precede all the other examples given here so far. ditional insight into the M = $ and M.= 7 patterns given here may be had by a comparison with African bel! patterns based on L= 16; anethes Sommon cycle length in West Alrica, (Otherwise, only & and occasionally 2 yeles are at all common.) In Table IT] a number of typical patterns are die layed despite the longer cycle length, M = S and 7 predominate here as vel, and all patterns except the last may be seen to contain exactly two.3's Prats ably the most widespread pater is number 1 (3324), whichis the basi cog beat considered to underlie virtually all the 4/4 Afto-Latin music of che Americas. Is basic collection of subunits may be used to derive patterns 3 an 3 by clement permutation, and all the other patterns by permutation and ‘ssion of elements, The clavé pattern may aiso be compared to the bese 2 stro ing ThisCOGNITIVE ISOMORPHISMS stroke pattern of 12/8 by aligning the two patterns synchronously and express- ing durations in units of 1/48th of a cycle the lowest common denominator. This yields, hil dd (ddl B+ Ro 1248+ 12/48 9494 1246+ 12/48, land it is evident that the two patterns partition the space in very similar fashion, despite their differing cycle lengths. In fact, begining Wester players often confuse the two, for their maximum time discrepancy (oceurring ai the 4th stroke) is of the order of 1/20 of a second at fast tempos. It may be readily shown by comparison with hypothetical alternatives (.g., 33334, 32434, 33244) that the clavé pattern is one of the two closest approximations 10 22323 that are possible in an L = 16 cycle (4/4 meter). The other pattern, 32434, is a permutation of the same elements. This kind of relation, which L ‘am going to call an example of analogue transformation, is a rule of corre- spondence between patterns of unequal L values, and may be represented approximately by coding the patterns in terms of elements of long (L) and short (8) duration, as has been done in Table II. This form of coding, though too imprecise for many purposes, may be useful for others. Singer, for example, in studying Macedonian dance steps, has profitably used such a representation, and it might he useful for scale characterization, ‘Analogue cransformiation between patterns may seem abstract, but itis not merely a theoretical supposition. African and Afro-Latin percussionists gener- ally recognize that some patterns from 12/8 and 4/4 meters may be used in tither rhythmic framework (e.g., C. K. Ladzekpo, Pertout). Furthermore, a fumber of the most clearly African-derived Latin percussion patterns exist in alternate 6/8 oF 2/4 versions, or have such versions of the same pattern sound ing simultaneous'y. This is found, for example, in the Cuban guaguanco and bara drumming styles, in Brazilian candomblé, and in the rhumba columbia (Pertout). ‘Next would like to look at examples for a very different 1. value, namely L = 7, This value is important because examples of pitch and rhythm structures rmay be readily compared. Here potential sources include, rhythmically, music of Macedonia, Bulgaria, Turkey, Persia and India, and in the pitch domain, Thai court music, music of the Chopi of south-east Africa, ‘Are'are music of the Solomon Islands (Zemp), and diatonic structures of Western music, Thai court music is known to be based to reasonable approximation on a tempered Senote system (Morton 1976), and this is also the tuning ideal for the Chop xylophone music. as revealed verbally ey the musicians and supported by averages of xylophone tunings (Tracey). The inclusion of diatonic Western suctares is based on the perceptual rather than frequency equivalence of the 7 tones in the scale, That this is so for the lay listener was pointed out to me by Roger Shepard, and is supported by research in Gestalt psychology (Koffka) ‘and categorical perception (e.g. Siege! and Siegel). ‘table IV lists the common 7-beat patterns of Balkan music and compares them to Western diatonic structures. The patterns given comprise all structures apee STUDIES IN MUSIC of fewer than $ sirokes which are composed of nearly equal elements of only 2 sizes (primarily 2 and 3 or 1 and 2; patterns of t and 3 are not common). A similar situation obtains with Indian falas. The isomorphisms between the two Spaces are again rather fer-teaching, since the given isorhythms define all the basic harmonic components of traditional Western tonal music: tonie= dominant relation and inversions of triads and seventh chords. The same prin- ciples of structural transformation found for the L = 12 case seem 0 hold here, as wel: eyclic permutation of pattern 2 (223) serves to generate patterns 3 and 4, and fission of the 3-clement into 2+1 or 1+2 can account for 5, 6. 7 and 8. Once again, the principle of complementation may equivalently be con- sidered to be at work, for the M = 3 and M = 4 patterns are complements of «ach other, as the reader may verify. Although Chopi anc Solomon Islands music seem t00 poorly understood at this time to form the basis for this sort of analysis, Thai music has been docu mented in some detail by Morton (1968, 1976). However, comparison with That music requires caution, as some natural criterion for idemtifying the unit cell pitch patterns needs to be invoked: there are no ‘chords’, for example, to unambiguously link sets of pitches. Typically itis found that melodies use $ of the 7 tones, or, in the mgm style, 6 tones or even all 7. However, these tones are ‘ot used with equal frequency or emphasis, nor are all used within a given Phrase. The importance of the Thai Sth relation is quite general (correspond. ing to pattern 1V—1), and the tones falling on the open and closed strokes of ‘the ching (small finger cymbals) are considered to receive special emphasis. A typical example js the piece Nok Khao Khamae (Morton 1968), where only tones 1, 3, $ and 6 occur on ching strokes. These may be seen to form pattern TW ~6 (2212), and a number of other pieces also yield basic patterns of Table IV. On the other hand, in the piece Khamén Saivdk, synchronous strokes of ching and gong emphasize first tones 1 and 6 (as in IV 1), and then tones 2, 3 and 6 (Morton 1968), This second case may be understood as a stacking of Thai Sths, generating a 313 structure which we did not encounter in the Balkans, though it has a pitch correlate in the common suspended sth chord of ontemporary jazz, So here the isomorphism is useful, but not as fat-reaching: itis possible that my limited wuderstanding of Thai music is at fault. (Metri, sally, Thai music is always 2/4 or 4/4, with the exesption of a body of songs in 1/4, with ching strokes divided 2212 [Moore]. INTERPRETATIONS: Group THEORY AND PERCEPTION It seems appropriate to consider why these few special patterns" have been Ghosen by so many apparently independent musical cultures, tis not an easy Question. The answer attempted here is based on simple mathematical moder, [ing and the stance of the cognotive psychologist; I have the belie that this Sommonality most tell us something underying’ about perception and the rin Perhaps the most elegant detivation of the widespread 22323 and 2212221 Patierns in the pitch domain has recently been given by Balzano, whose ideas Bex Meinal for the interpretation offered here. While acknowledging otter special historical sources for our tonal structures, such 2s vibrating body overs “ tones, proper from examp ‘wo sn reducit ture as x43 Anim diffe from rot ge same trast, prime group 1 jus 7 gene (the L bole eur Itt thoug Tare the te the fu comm The Altho Pieces or 332 Latin be cor take unit ment, nit i This used | Ciagt For ony or |f Patter tione this soma on asCOGNITIVE ISOMORPHISMS tones, the tetrachord and the desire for modulation, he finds the special properties of the pentatonic and major scales in our 12-note system derivable from mathematical group theory. In particular, our musical system is an example of the abstract group C2, and itis expressible as the direct product of ‘wo smaller subgroups as Cs * Cy (@.g., Budden). A consequence of this reducibility to subgroups is the natural representation of the major scale struc- ture as a convex region in a two-dimensional lattice with unit cell dimensions 3 X 4; 3 and 4 semitones are the pitch distances used to define triads, of course. ‘An important and natural feature of this group theoretic representation is the differentiation (for L = 12) of patterns based on the distances 2, 3, 4 and 6 from patierns of other lengths, since these numbers, being factors of 12, will not generate all elements of the group if replicated, but rather reproduce the same (6, 4, 3 or 2 respectively) elemenis after the distance of one cyele. In con- trast, the numbers 1, 5, 7, [1 are group generators of Cy2, because they are prime with respect to the group order (12), and will yenerate all elements of the ‘group before repeating any (Boretz 1971). In musical space, the numbers 1 or 11 just generate the chromatic scale or underlying rhythmic pulse, while $ and 7 generate the cycle of Sths: four Sths generate 22323, while six Sths generate (tie Lydian mode of) 2212221. In rhythmic terminology, polyrhythms are built up from subgroups, and time-lines are formed by replicating a non-unit generator, If this same procedure is tried for L = 7, similar results are achieved, though here the group order (7) a prime number so that all numbers less than 7 are generators. Since there are no subgroups there are no polyshythms, and the replication of 2, the smallest non-usit generator, generates 223 and 2221, the fundamental structures for this size cycle, Structures generated by 3 are not ‘common in the musics under discussion. ‘The case of L = 8 may also be mentioned at this point for completeness. Although pitch space data seem to be unavailable, rhythmically such Ewe pieces as Kinka and Sovu, and the Lobi popular music Nandon, utili or 332: the samba, habanera, tango and innumerable other African-influenced Latin American dance forms use it as well, (Although two of these cycles may be considered to form a cycle of 16, 1 have not done this here, preferring to take the smallest recurrent unit as fundamental.) In any case, the smallest non- unit group generator is 3, and generates precisely this pattern. Its comple- ment, J $4 | may also be generated by 3, and is the basic replicating tnit in the 24-eyele bell pattern of the Dahomey piece Kadodo (Combs 1975). This complement is also widespread in Afro-American music (Evans), and is used with displaced starting point as time-line for the Ewe piece Adzro (Fiagbedzi) and the Ashanti piece Kumachacha (Combs 1974). For L = 16, where again we have only rhythmic information, the smallest non-unit generator is again 3, which yields (on 4 replications) the pattern 33334 or PG f ‘DJ |, which is almost but not quite the fundamental clavé patlerns The difference is slight, for 4 of the $ strokes are identically posi- Toned, and the other, the ath, differs by only one semi-quaver.'* Nevertheless this discrepancy poses a problem for the group theoretic derivation of pattern commonality offered here, The only plausible explanation seems to be based fon ascribing this difference 10 the perceptual pull of the predominant 12 4sgars STUDIES IN MUSIC pattern 22323, using the idea of analogue transformation presented earlier, ‘That is, the ‘intrinsic’ pattern 33334 has been stretched siightly to become the 33424 clavé chythm on the basis ofthe latter's close perceptual analog relation Ship to the widespread 22323 pattern. This explanation in terms of underlying cognitive structures and simple group properties might be expected to apply to aspects of the music other ihan bell patterns and diatonic scales. For L = 12, where this possibility can be ‘most clearly tested, it does seem to hold, The basic African 2-, 3, 4 and 6-clap cycles are respectively equivalent to the tritone, the augmented iriad, the diminished 7th chord, and the whole ione scale, In both pitch and rhythrn space, the 4-cycle is in some sense the most fundamental of these possibilities: the 4-clap constitutes the ‘basic pulsation’ (C. K. Ladzekpo; Locke), while the diminished 7th chord is @ mainstay of ‘modulation for tonal music. Extensive use of 3- and 6-clayy patterns as well in African music points out that its best pitch analogs are late romantic, impres- sionist, early 20th century music, and of course, contemporary jazz. In facts Just the Seales used in jaze (Pressing) are particularly well represented in West African shythm space, as may be seen in Table V. Since pitch and rhythm domains are subjectively so different (serialist music notwithstanding), it is Perhaps surprising that these structural analogies should go as far as they do. The differences that do exist provide some interesting information. For instance, in rhythm space the major triad structure does wot turn up promi- rently for L = 12 as it did for L = 7, This seems to support the idea of the triad asa diatonic seale-derived concept based on synchrony; it may also be duc to the low filing ratio value this corresponds to (see below). And two Scales occurring in jazz which do not, to the best of my knowledge, correspond to common West African patterns, throw additional fight on the matter, Har- monic minor (2122131) and harmonic major (2212131) scales are the least common jazz scales. They are the only scales here which have clements of three sizes (1, 2, 3); hence they are in some sense less economical in their use of ‘material. and sample the space more unevently. Also, each of these contains the diminished 7th chord, which corresponds to every an-beat of some under {ying basic ¢-pulsation. Since the time-lines arc already perceived in composite ‘elation with the underlying 4-pulse (Locke: C. K. Laczekpo), these patcerns may lack the peteeptval flexibility or variety required of a maximally versatile fime-line." in fact, one of che interesting properties of the 22323 and 2212221 Fuclures, not shared by any other structures here, which supports such an idea, is theit property of sampling as equally as possible the various sub Giisions (subgroups) of 12: 2, 3, 4and 6. Mathematically this occurs because the pattern (group) generator S (or 7)is prime shares no common factors) with {especl 10 these numbers. For instance, there are only 2 distinct G-cycles, which interlock, as indicated by the following pattern of x's and o's: cycle 1 SekoKOKORONO cycle 2 occurs f tively; 24242 14141 equal s very gr based 0 propert whieh shythmi sciously percept (see Lox This: standi ratio, i one wo sample tions." by Rah will of forL = 16,9 uite st the cor analogt the L Adzoh Ano iseven even n 2-94 not itm havea (especi Attica Aud posed unless difere made.COGNITIVE ISOMORPHISMS (In pitch space these are the two complementary whole tone scales.) The 22323 and 2212221 structures, regardless of at what point they are begun, when over- laid on this, divide their strokes as nearly equally as possible between the x-and o-cycles; ¢.8., 3a) + 2(0) and A(x) + Ho), respectively. That is three strokes of the 22323 patcern will fall on elements of one 6-cycle and two on clements of the other 6-cyele; four strokes of the 2212221 pattern will fall on elements of ‘one 6-cycle and three on elements of the other. A similar equal partitioning ‘occurs for the three 4-cycles: xo —xo - xo—x0~; 242+] and 24243 respec. tively; and the four 3-cycles: x+0—x+0-x+o-; 2+1+I+1 and 2424241. And all six 2-cycles produce the respective sampling patterns 14141414140 and 14141414142, The musical usefulness of this equal sampling distribution generated by the fundamental patterns may be very great. Suppose the African master drummer wishes to play @ pattern based on one of these cycles, with specified starting point. This sampling property means that there will always be several bell strokes he can tune into Which will not only locate the proper starting point, but generate the correct rhythmic feel as well. I do not know whether such a process goes on con- sciously, but it must facilitate the aggregation of underlying pulsations into perceptual units of various sizes, which is so typical of master drum patterns (ee Locke; Jones 1959). “This idea of equal sampling may be used in another way as well—in under- standing the choice of M for a given L. Lets = M/L define a rhythmic filing ratio, in the manner that von Hoerner has done for pitch. Then, as in his work, ‘one would predict that this number should not be too far from ¥4, 50 as to sample the space adequately, and make maximal use of figure-ground rela- tions,’ Such a prediction also follows from the theory of bisection proposed by Rahn (1977, 1978a). In practical terms, this means that elements of size 2 ‘will often predominate. And the most common patterns looked a: here yield for L = Ides = 8/12, 1/12; for L = 7,8 = 3/7,4/7: for L = 16, 5 = 5/16, 7/16, 9/16; and for 1. = 8, s = 3/8, $/8. These numbers support the idea quite strongly, except for s = 5/16, which seems to confirm the specialness of the corresnonding clavé pattern and support its derivation from 22323 by analogue transformation, Two other bell patterns which support s = ¥% are the L-= 24 Ago (s = 11/24) and Kododo (s = 15/24), part of the suite ‘Adzohu, from Dahomey (C. K. Ladzekpo; Combs 1975). “A noticeable feature of these s = J filling ratios is that, particularly when L iseven (12, 16, 8), M is almost always odd; notably, the exact middle value, the ‘even number L/2 seems to be avoided, except in some Ashanti music (pattern 3°"9, and even there the 6th stroke is often omitted). Why this should be so is not immediately apparent from the group theory explication given heres it may have 2 base in aesthetic preferences or special meanings given these numbers (especially § and 7) in these cultures (cf. Zastavsky for special properties of African number systems). ‘At this point a look at the constraints, both cognitive and kinesthetic, im- posed by the physical act of performing these shythms seems appropriate. For, JBnless all the constraints imposed on the basic perceptual space correspond in Uifferent representations, itis unlikely that similar structural choices will be made. 0STUDIES IN MUSIC It seems quite significant that the tempos of West African and Bulgarian or Macedonian music are all relatively fast. Nyayito, one of the very slowest Ewe pieces, moves at something like ) = 226, and tempos for fast pieces like Atsiabeko are. } = 600+, close to the possible performance limit. Likewise, Bulgzrian and Macedonian tempi are quick, and if } is taken to be the basic fastest pulse, range most commonly from = 250-500. Considering these speeds with respect to the finger, writ and arm movements required in perfor- ‘mance suggests that the quickest pulse functions both as a cognitive grid and a kinesthetic threshold, which may be decorated occasionally in the service of heightened rhythmic activity, but whose underlying basis cannot be easily controverted."* This kinesthetic threshold facilitates the cyclic time-line idea in the same way that instrumental design and physical playing habits reinforce any but the most determined Western microtonalists from making subdivi- sions of 12-tone temperament more than occasional features of theit composi- tions. And likewise, fast tempos would be expected to limit structures com. posed of strings of 1's, In fact sequences of 1's longer than 2 are not found it Macedonian dance (Singer), and even adjacent I’sare absent from all the basic African time-lines, (The Ewe Nyayito, which is a rare exception, is compara- tively very slow and may be considered a composite pattern, as noted earlier.) This same property is found mirrored in the pitch domain for different (har- monic) reasons: Pressing has shown that the avoidance of adjacent hralf-steps isa principle which underlies the construction of jazz scales. Fast tempos also facilitate the focusing of attention on the cyele as a whole, rather than only on serial relation; again, this fact argues for patterns with holistic group transformational properties. World musics with characteristi- cally slow tempi and cycles of 12 do not, in fact, use the basic time-lines found here ‘Consider, for example, the basic kutkéri pattern of Korean music, gidubbodd ara T Tp gisddaded ortaryong, *f © PF * ~ stayed on both ends of the changgo drum (Song). The top lines here are respectively 221322 and 321231; the slow tempi do not favour polyrhythms, and these patterns are unlike those found earlier. A similar phenomenon obtains with characterstically slow Turkish music, with patterns like a dN PHY oad Med Fy @arban And African melodie phrase lengths, considered in this cyclic fashion, are not dominated by the divisions of 12 given here, though they do oecur (see Ewe children’s songs, Jones 1959), ‘This suggested dependence of pattern on tempo is supported by considerable "esearch in cognitive psschology which finds different kinds of perstpeeal organization operating on different time scales, This has been conceived can, linking up of internal vine scales withthe time scales of extanal events ap described in the work of Mari Jones. Other examples ae the demonsiratel i 48 portane 1972), and Hi exampli at slow fast tem traditio Ment 1978), tural sir of near native ‘number efficien is less t Also ge tive psy portant and Bu 1976), Expe here's ing rep torepr of vary and gr startin the pat this for mute s More s prefert unders clemen eval. percep ing poi were se sible p Predict 2,COGNITIVE ISOMORPHISMS portance of tempo on the perception of single line rhythms (e.g., Preusser 1972), and also on the perception of polyrhythms, as reported by Oshinsky and Handel (1978) and Handel and Oshinsky (1981). They found, for ‘example, that 2 x 3 polyrhythms were perceived predominantly as triple meter at slow tempos, but predominantly as duple meter or a composite pattern at fast tempos. How such findings might differ for African listeners, with their tradition of composite 2 versus 3 hearing, is an interesting question. Mention should be made at this point of the work of Rahn (1977, 1978, 1978b), which can provide the basis for an alternative explanation of the struc- tural similarities presented here. His theory of bisection allows the generation of nearly all the patterns given, and must be carefully considered as an alter- native. The only disadvantage which may be pointed to is the fact that the number of possible patterns it generates is far greater, and hence its predictive efficiency in explaining these particular widely occurring isomorphic patterns is less than that of the current group theoretic derivation. PERCEPTION OF TEMPORAL STRUCTURE ‘Also germane to the present theory are the results of recent research in cogni- tive psychology on the perceptual organization of temporal patterns. Here im- portant contributions have been made by Fraisse, Royer and Garner, Handel and Buffardi, Preusser et al. (1970), Handel (1973, 1974), Restle (1970, 1972, 1976), Garner (1974), and Handel and Todd (1981), among others, ‘Experimentally typical and of particular relevance to the theory presented here is the work of Garner and co-workers, Their experiments entailed present- ing repeating sequences of dichotomous elements (0 subjects and asking them to reproduce the patterns. Most of the work was done with 2-element patterns of varying lengths, with the elements represented by high and low tones, ot red land green lights. The information obtained included the preferred perceived Starting point(s) of the recurrent structures, and relative difficulty of learning the patterns as measured by the number of cycles required to learn them. In this form the data might be applied most directly to the two-toned bounce and mute strokes of the Ewe drums kid! and sogo, but this will not be done here.’ More significant for the time-line structures of this paper is the result that the preferred pereeptual organizations of the 2-element patierns proved to be linderstandable in terms of organization of the constituent complementary 1- Clement patterns, a representation equivalent to that of the timerline (Preusser etal. 1970; Garner 1974), Basically it was found that the difficulty of pattern perception corresponded closely with the number of preferred perceived start- Ing points in the endless cycle presented, and that runs of successive elements rare seldom, if ever, broken up (Royer and Garner 1966). Furthermore, pos- ible perceived starting points (hence perceived difficulty) could be accurately predicted by the interaction of two principles: 1. Pattern descriptions tend to end with the longest gap between elemenis (gap principle): 2, Batter descriptions tend :o begin with the longest run of elements (run principle). 0STUDIES IN MUSIC When he two princises agree, learning the patter is easy and unambiguous; when they disagrce, pattern learning is more difficult, and a variety of starting, Points is perceived. For example, consider the clavé rhythm, 33424: there are no Tuns of adja- cent elements ( FZ), so that principle 2is not operative, whereas the gap prin- ciple predicts that the pattern should end with one of the two 4’s, as it does, ‘The application of these principles to time-lines requires some caution, how- ‘ever, for, first, 4 composed pattern can be learned with any starting point; second, it is possible that these Psychological data are culture-bound; and third, the tim is part of an ensemble and not the only important aural stimulus, In addition, as mentioned above, ‘Many time-lines given here are con- ceived in relation to an underlying pulse, each beat of which spans 3 or 4 smallest time units; because of this additional Perceptual constraint, these rhythmically divisive instances are less completely isomorphic to the psycho- logical experiments under discussion than are cases of purely additive organ- ization (¢.g., Balkan chythm, Western scales), Consequently, while many im- Portant time-line structures here do follow the rules postulated above (¢. 22323, 23223, 33424), many others, notably those with Preparatory upbeats (€.g., 2212221}, do not, and the issue may be further complicated by che philo- Sophical idea (as among the Ewe) that the start and end of a Pattern can be the same Examination of the time-lines given here seems to indicate that an extension ‘of Garner's two principles to account for the possibility of perceptual aggrega- tion is required, especially in musical context. Such perceptual aggregation is common knowledge among performing musicians, and has been described fecently iu some detail by Handel and Todd (1981). Specifically, perceptual subunits may be construed, facilitated by the organization of the pattern and an underlying pulse (if present), which consist of onc or more elements and ‘one or more gaps: the additional principle Proposed here then states that Pattern descriptions tend also to avoid breaking up runs of such Perceived ‘aggregate subunits. For example, in this way of thinking the basic pattern 22323 may be considered to be made of unequal ‘elements’ of 2’s and 3°s, rather than single elements and gaps. This is facilitated by the complete. absence of 1’s from the pattern, and would provide the prediction that 4 of the 5 modes should be perceptually well organized: only 23233 would bresk the single run of 2-elements and hence be a Poor time-line choice, As the reader will recall, just this situation obtains for the West African and Balkan musics here, Likewise, the preferred Starting point of Ashanti pattern 1-8, 2121222, {s Feadils understood from perceptual aggregation as 2121222, with the tan of 2's and the run of geregated 3's remaining unbroken. imilarly, the invoked third principle may be considered to explain the rarity of 232 in Balkan music, andis consistent withthe preferred starting points of all the L. = 16 patterns of ‘Table ITl, which do not follow Garner's first two principles alone, The only remaining patterns whose starting points require further explana- tion ate 2212221 and 1222, where in each case all modey may in Principle occur ‘but those beginning with 1 are rarest (compare phrygian and locrian modes, Seeaale WD * The following explanation canbe offered in light of prince 3: both sequences are dominated by runs of 2's which are broker up, or rather so shifte lem quick explan leadin Are of Ro} elemer patten terns dent a most xoxo be che are te duced Note samp stabil gener shown third need f from perce} shyt (ee, simple will al ing fo and p play aCOGNITIVE ISOMORPHISMS shifted, by the I's. To have the start of a cycle begin with 1, the short shifting clement, without a possibility of perceptual aggrezation, is jarring enough (at quick tempos) 10 make those patterns rare as reference structures. (An explanation in pitch space would likely need to invoke principles of tonality or leading tone.) A related perspective on perceived patiern difficulty may be seen in the work ‘of Roger and Garner (1966), who carried out a detailed study of all possible 2- clement patterns of length 8; unfortunately. no comparable examination of patterns of length 12 has been reported. For L = 8, although 2° = 256 pat- terns are in principle possible, only 16 2-clement patterns are actually indepen: dent and not reducible to smaller eycle lengths. Of these. 1wo were by far the ‘most perceptually ambiguous and difficult 1e learn for subjects: wwoxxoxo and xxoxooxa. In terms of the constitutive I-element patterns (either x's of o's may bbe chosen as the foregound, with the other symbo! supplying the gaps), these are respectively 12122 or 332, and 1232, and are exactly the } patterns pro- duced hy the Cs group generator, 3. That is: °—332, #~1232, #~ 12122, Note that s = ML =‘, Because these patterns may be so generated, they sample the 2- and d-cycles af this group uniformly, and their perceptual multi- Stability may be considered co reflect their unique derivation from the group generator. The first two parterns, as 332 and its complement 12122, have been Shown to dominate West Afrivan (and Afro-Latin) 8 « yortayado ECITNZ “6 sav oa aaa (weuoc zw) TZLIIT “8 gy pia go weusoy = TZZIZZI “ek oa gaa uehjoy zeizeiz “9 voaaa MimpsioniIn, zvezize *S avowigaa uepaAyT 1271772 OP a yor a <> uerbAiyd = T2727 savo4 aad ueuOg = ZITeeIZ T avoaaao CT PE Pf ewop ayes soleus recite 1 spouaia}aa uouejou POU ISON (Qupuopeiow — woned wos) s3}durex> urewop ue aureu wana urewop youd (souyj-suin) wIAYL pu (S9]895) TaTavi id 203 suaned Z{ = 7 °2 = W Jo vosueduioD, 3STUDIES IN MUSIC UE a iove (6s60 so00p (onsesn) amet odyaepet “4 ‘9 (90016 Guu) meysy moopeiueg (warmed puw odyaepo1-3°s asimve MpeyR) 263, (E96) BNDAN dy) BD (p61) squi0e> femopy) ueYy (sun20e (961) s9u0r waned ew) equiog surponig «(Sus s,uospn9) epus, (£960) RODIN Q@deieyy) uryy eurersy *(@rE961) BN24N, (5961 *6S61) sauog Sun YS fodysepen "MD SuayIO KURU Fy) 28g (e961) 2H>1N (@uos susan Heyy (661) squio. ‘2unn204 waned nuEysy Yom ssouniajos EDLY 152. wor soydunex CP MC f avorssavaas ter tr quvoa(@> Trerre veqqao terry tT Wot ga rrr t sand go fret 4 gd ga rrret voad>. rreorere voaa2 trent wpe arte Tt wag uoneyou ——uorreiuasaidar urwaiop wud wynip (l= 7 %6°9'S “e = Wy siuseambs youd aroys pur sussned 1199 8/21 2340 uatava jou wo dyouIs arm | uioned deo reer iz ziceee eezez pzize zeece zeeze eceez exeze weer veze 6 or 96 se Card rol usoned w $8COGNITIVE ISOMORPHISM (SLO1 “P61 Sat.» odyorpe 14>. odyoepe joy pure ny 4 Pa spouianayas aulduepy (20421) auiBuEpYy-eD (esird aurduepy-1y IN O8e Sayer) 2M yrag 94RD use pousy 243, Ssuauepy-ry) “UeYV, wou BoA. uaosy sop susmiged qq URTV SAK OL = IsIsTITeT “ISSISss AS ISsss SSSS1ST sss STISS Buoy- ous me arrava urewiop wey rea dyons anu f waned dep , eicieeeee eeeeree eeereze eeeeeze every thee ropes ond 6% a) Ls LotSTUDIES IN MUSIC spaoya yiuasas pue spein jo sod) VOWNUOD {IF 1FYI OS *pasn 9g jYBIUL >POL d1UCIEIP AUE ING “aB>S LOFE UO 949y PIs (vonesrunui0> Teuossady “24109 "a roauig souuauaaay zequrg soqqua souqua 434 214 wU93y wsy aay9MoU) ‘21st wersiag uF sun930 tumouy auou oyfadery ood sWoW OU2TF eznusysn ina Sueqieg ut ove 9104 oxsuopaynyy, eriuayrns (ous UPA UPYIPEL wwyesys pur yp op yay pid say al aia puoyp yiuanas vorssnaue pag WAC? puoy yruanos uoissonut pug ‘Vy. vorsiaaut is] WORD anys qnuaaas Wod 1004 5493.) pen wouaRU IST WAD pew uorupaut puz (V4 peu boned 1001254.) ‘pannus aq ues wtb we way ize wz we etuommussaadas afeas ur wey 9494 pest mod SurseIs 14929534 Loy Tepow us yursaid (6561) sou0r ssa0nid 9 - odyazpe 1 4-2 Kuowur edey CPCCA aves yd gd dD ztCIZITL 8B COM NTE ayostas wine 6 CIM | a yodaao ae ok znpasy ur 14 odyarprq “4-9 en syvurasexe C fF PP ty tT AVOd GAGS —12TIT Los waaned 218e9 Laqqn 398 n syceepeaty ¢ tf te tt AVDIIGD —UZCLATT c+ (6561) savor suuaiyed ayo118 sodxarpe yD sounoqiprvana TEP TES TEE wosaga> ell (6561) s9u0r ops prt ttt werowsaad wee aiae 25 wamedosea ff oT TT voaa> eee s900939494 ‘Due wapambo Saye ae wed wo) s9yduers anus surstydsowost uyncya atop pur (BuIsso1g) sampanus 91295 Z7eF WOWUIUOD AqL, A aTEVL “parruauad aq ues 6